What is this, the third or fourth time this same exact RFC has been posted today? Give me a break. One of Taco's friends must have written it. The whitespace-based programming language was much better.
If anyone could actually read the RFC (since it's been slashdotted multiple times and fainted from all the exertion), you'd see it only applies to TCP over carrier pigeons anyway. Who uses that anymore? Savvy network administrators run IIS with SCTP tunneled through those little cameras they put on baby whales.
...there is no way in hell I'm installing 5.0 on anything important, even though it's going to be a "production" release. 4.8, 4.9 all the way baby.
Why you ask? There's far too much new code for 5.0 to be stable yet. I was using 5.0-CURRENT SMP in November and December of 2001, and was very impressed. Alas, it was running on an IBM DeathStar 75GXP, which died (lol--like the name suggests)...
I unsubscribed from the -current list a month or so later because Matt Dillon (the real one) was being his usual dickheaded self and causing a massive flamewar.
Anyway, I resusbscribed to -current in October cause I knew they had slipped the release date to somewhere around November, January, etc. and I wanted to find out how things were going (i.e. is this good enough that I should install it and have more fun with it). Ooh boy. Since I left, we've added GEOM, GDBE, a new init script system, IPFW2, UFS2, etc. vn has been replaced by md, devfs hasn't gotten any better, and as far as I can tell, they still have background fsck turned on by default, which tends to hose you when the least thing goes wrong with your fs (background fsck was FreeBSD's bitter parting shot to me when my GXP died -- it murdered my filesystem before I had a chance to save my valuable data -- admittedly this was a "for fun" desktop system, but that's typically considered Naughty). On -current today we have a couple people posting about panics. I enjoy the response in this one:
From: phk@freebsd.org (for those who don't know, Poul-Henning Kamp is one of the wisest, most respected, and ancient of all FreeBSD hackers)
Subject: Re: FreeBSD 5.0 RC3 now available
"Roderick van Domburg" writes:
I would like to point to a currently unresolved issue
[snip]
The thread is titled "panic: trap: fast data access mmu miss" and is about an error causing the sym SCSI controller to fail to mount root at best, and panic at worst.
Mr. Henning-Kamp's response:
Well, we all want our pet bug fixed before the release rolls, but at some point we simply have to call it quits and ship the release.
[snip]
In the meantime we _really_ have to ship 5.0-RELEASE, we keep slipping it.
Commentary: I agree, they really need to get 5.0 out the door, and I don't necessarily disagree with phk's opinion. But it does say massive Bad Things(tm) to me about the quality of this software that release engineering is leaving *known panics* in the software cause it is so late and over-schedule!!! Ah, and don't even get me started on not being able to install new boot blocks or run fdisk on a mounted filesystem, crashdumps overwriting people's disklabels, etc. etc.
Another one just came in: "PANIC in tcp_syncache.c sonewconn() line 562" about an easily-reproducible (from user mode) kernel panic. Come on people, this is worse than Windows NT ever was! (well, except the guy who could bluescreen it by printing tabs and backspaces).
Slashdot Mirror
What is this, the third or fourth time this same exact RFC has been posted today? Give me a break. One of Taco's friends must have written it. The whitespace-based programming language was much better.
If anyone could actually read the RFC (since it's been slashdotted multiple times and fainted from all the exertion), you'd see it only applies to TCP over carrier pigeons anyway. Who uses that anymore? Savvy network administrators run IIS with SCTP tunneled through those little cameras they put on baby whales.
...there is no way in hell I'm installing 5.0 on anything important, even though it's going to be a "production" release. 4.8, 4.9 all the way baby.
Why you ask? There's far too much new code for 5.0 to be stable yet. I was using 5.0-CURRENT SMP in November and December of 2001, and was very impressed. Alas, it was running on an IBM DeathStar 75GXP, which died (lol--like the name suggests)...
I unsubscribed from the -current list a month or so later because Matt Dillon (the real one) was being his usual dickheaded self and causing a massive flamewar.
Anyway, I resusbscribed to -current in October cause I knew they had slipped the release date to somewhere around November, January, etc. and I wanted to find out how things were going (i.e. is this good enough that I should install it and have more fun with it). Ooh boy. Since I left, we've added GEOM, GDBE, a new init script system, IPFW2, UFS2, etc. vn has been replaced by md, devfs hasn't gotten any better, and as far as I can tell, they still have background fsck turned on by default, which tends to hose you when the least thing goes wrong with your fs (background fsck was FreeBSD's bitter parting shot to me when my GXP died -- it murdered my filesystem before I had a chance to save my valuable data -- admittedly this was a "for fun" desktop system, but that's typically considered Naughty). On -current today we have a couple people posting about panics. I enjoy the response in this one:
From: phk@freebsd.org (for those who don't know, Poul-Henning Kamp is one of the wisest, most respected, and ancient of all FreeBSD hackers)
Subject: Re: FreeBSD 5.0 RC3 now available
"Roderick van Domburg" writes:
I would like to point to a currently unresolved issue
[snip]
The thread is titled "panic: trap: fast data access mmu miss" and is about an error causing the sym SCSI controller to fail to mount root at best, and panic at worst.
Mr. Henning-Kamp's response:
Well, we all want our pet bug fixed before the release rolls, but at some point we simply have to call it quits and ship the release.
[snip]
In the meantime we _really_ have to ship 5.0-RELEASE, we keep slipping it.
Commentary: I agree, they really need to get 5.0 out the door, and I don't necessarily disagree with phk's opinion. But it does say massive Bad Things(tm) to me about the quality of this software that release engineering is leaving *known panics* in the software cause it is so late and over-schedule!!! Ah, and don't even get me started on not being able to install new boot blocks or run fdisk on a mounted filesystem, crashdumps overwriting people's disklabels, etc. etc.
Another one just came in: "PANIC in tcp_syncache.c sonewconn() line 562" about an easily-reproducible (from user mode) kernel panic. Come on people, this is worse than Windows NT ever was! (well, except the guy who could bluescreen it by printing tabs and backspaces).
So, no thanks to 5.x for me, for now.